Forschungsthemen
[MA] Efficient, Remote, Parallel Execution of SLAM Algorithms on Kubernetis
In previous work at the Software Technology Group, the GeneralRobots framework was adjusted, to allow fort he remote execution of a selected part of a particle-based SLAM algorithm. The hypothesis was that it is possible to reduce the overall efficiency of SLAM by decomposing the SLAM algorithm and executing the resulting parts on a remote machine in parallel. The goal of this thesis is to investigate how the SLAM algorithms given by the GeneralRobots Framework can be systematically decomposed aiming at their efficient, remote, parallel execution. For this, an existing Kubernetis infrastructure shall be used.
The following tasks have to be fulfilled:
- 1. Get familiar with the GeneralRobots framework (https://git-st.inf.tu-dresden.de/haec/slam-haec)
- Find and compare to related work
- Find literature on the general architecture and decomposition of SLAM algorithms
- Find and compare to literature on „cloud-based SLAM“
- Provide an architectural overview for
- The GeneralRobots Framework
- The SLAM Algorithms in the GeneralRobots Framework
- Develop a concept for decomposition of SLAM algorithms aiming at their efficient, remote, parallel execution
- Implement the concept using an existing Kubernetis infrastructure
- Find/select a benchmark and evaluate the implementation
Betreuer: Sebastian Götz